Week 1 Flashcards
- Which of the following best defines Artificial Intelligence (AI)?
A. A robot that replaces human labor
B. A machine programmed to act randomly
C. The ability of a computer or machine to mimic intelligent human behavior
D. A form of software used in hospitals
Correct Answer: C
- What distinguishes Machine Learning (ML) from traditional programming?
A. ML requires no data
B. ML is manually coded for each task
C. ML enables computers to learn without being explicitly programmed
D. ML uses less memory
Correct Answer: C
- In healthcare, how does AI contribute to fraud detection?
A. Replaces doctors during operations
B. Provides spiritual support
C. Detects anomalies in billing and patient claim patterns
D. Trains staff
Correct Answer: C
- Which of these is a key application of AI in the power sector?
A. Enhancing student grading
B. Managing social media posts
C. Predicting energy demand for smart grid optimization
D. Managing parking lots
Correct Answer: C
ML is a subset of AI.
True
Generative AI can be used to create synthetic medical data.
True
AI cannot be used in art or entertainment because it lacks creativity.
False
AI-powered systems are unable to support mental health therapy.
False
A hospital wants to implement an AI system that predicts patients at high risk of heart disease using demographic and health data. Describe a basic workflow for this.
Solution:
Data Collection: Gather patient data—age, weight, cholesterol levels, blood pressure, smoking history, etc.
Data Preprocessing: Clean and normalize the data.
Model Selection: Choose a supervised ML model (e.g., Logistic Regression or Random Forest).
Training: Use historical labeled data (high-risk vs low-risk) to train the model.
Evaluation: Test on unseen data and measure accuracy, precision, recall.
Deployment: Integrate with hospital systems for real-time risk scoring.
Create a simple example showing how an ML model could be used to automate prior authorization in health insurance.
Solution:
Input: Past claims data (patient ID, diagnosis, treatment, approval status).
ML Task: Binary classification (Approve/Reject).
Steps:
Encode textual data (diagnosis, treatment) into features.
Train a Decision Tree classifier.
When a new claim is submitted, the model predicts approval likelihood.
Why do you think AI adoption in healthcare must be handled with caution despite its benefits?
Sample Answer:
While AI can enhance diagnosis, reduce costs, and improve care, it raises concerns around data privacy, algorithmic bias, lack of transparency in decision-making, and over-reliance on machines. For example, an AI model trained on biased datasets may produce unfair recommendations, potentially harming underrepresented groups. Thus, ethical and legal frameworks must guide AI integration in healthcare.
How can generative AI help solve the problem of limited medical image datasets in developing countries?
Sample Answer:
Generative AI can create synthetic medical images that mimic real ones without exposing patient data. This enables better training of ML models while preserving privacy. In regions with data scarcity, it helps researchers access diverse datasets to improve diagnosis models.
Scenario:
An HMO wants to allocate flu vaccines more effectively. They want to use AI to detect early flu outbreaks across different states.
Q: How would you apply AI/ML to solve this problem?
Step-by-Step Approach:
Data Collection:
Pull real-time data from clinics, pharmacies, Google flu trends, weather patterns, and hospital admissions.
Feature Engineering:
Extract features like number of flu cases per region, season, patient demographics, and temperature.
Modeling:
Use time-series forecasting (e.g., LSTM) or classification models to detect outbreak patterns.
Prediction & Action:
Predict likelihood of outbreaks and recommend vaccine distribution accordingly.
Outcome:
Early interventions in high-risk zones → reduced hospitalizations and efficient use of resources.
